Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session B: Wednesday, August 13, 1:00 – 4:00 pm, de Brug & E‑Hall

Inferring Cognitive Load from Deviations in Simulated Human Planning Behavior

May Kristine Jonson Carlon¹; ¹RIKEN

Presenter: May Kristine Jonson Carlon

Understanding how cognitive load shapes human planning behavior is crucial for building AI systems that collaborate effectively with people. While traditional approaches to measuring cognitive load such as self-report questionnaires or dual-task paradigms are valuable, they often lack real-time responsiveness or introduce artificial task constraints. This work is a proof-of-concept for inferring cognitive load from deviations in planning, thus avoiding intrusive or retrospective measures. We simulate user profiles performing a structured task (summarizing an article), with behavioral noise introduced via repetition, backtracking, pausing, and skipping actions. A Hidden Markov Model (HMM) is used to infer latent cognitive states from the resulting behavioral traces. Results from 100 Monte Carlo trials show that the HMM reliably recovers latent states aligned with intuitive levels of cognitive load. Emission patterns are interpretable, stable across trials, and distinct for each state, capturing predetermined behavioral signatures of low, medium, and high mental effort. State assignments also show alignment with simulated user profiles. Our approach provides a simulation-based foundation for modeling cognitive variability and may inform future work in user modeling, Theory of Mind, and adaptive systems.

Topic Area: Reward, Value & Social Decision Making

Extended Abstract: Full Text PDF